Artificial Intelligence - Page 11 - E46Fanatics E46 BMW Social Directory E46 FAQ 3-Series Discussion Forums BMW Photo Gallery BMW 3-Series Technical Information E46 Fanatics - The Ultimate BMW Resource BMW Vendors General E46 Forum The Tire Rack's Tire Wheel Forum Forced Induction Forum The Off-Topic The E46 BMW Showroom For Sale, For Trade or Wanting to Buy

Go Back   E46Fanatics > Everything Else > The Off-Topic > General Off-Topic

General Off-Topic
Everything not about BMWs. Posts must be "primetime" safe and in good taste. You must be logged in to see sub-forums.
Click here to browse all new posts.

Reply
 
Thread Tools Search this Thread Rating: Thread Rating: 5 votes, 5.00 average. Display Modes
Old 01-13-2017, 10:10 AM   #201
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Now poker isn't safe
https://medium.com/@Moscow25/ai-vs-p...ead#.vp62maq8w
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 01-26-2017, 12:52 PM   #202
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Next up: X ray techs and radiologists
https://www.wired.com/2017/01/look-x...ai-coming-job/
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Last edited by cowmoo32; 01-26-2017 at 12:52 PM.
cowmoo32 is offline   Reply With Quote
Old 01-30-2017, 07:34 AM   #203
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Key finding is 1:23-1:46
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Last edited by cowmoo32; 01-30-2017 at 07:34 AM.
cowmoo32 is offline   Reply With Quote
Old 02-15-2017, 10:23 AM   #204
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Microsoft just released a simulation environment that uses Unreal engine and allows for testing of autonomous vehicles and other machine learning...stuff....
https://github.com/Microsoft/AirSim

__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 05-08-2017, 06:02 PM   #205
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Watch a neural net drive in GTA V
https://www.twitch.tv/sentdex
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 05-08-2017, 06:28 PM   #206
GasKing
OEM ///Member
 
Join Date: Oct 2012
Location: Tri-State
Posts: 848
My Ride: 18 Tahoe RST
Stimulation environment. Yeet.


Sent from my iPhone using E46Fanatics
GasKing is offline   Reply With Quote
Old 05-18-2017, 10:38 AM   #207
jeffro3000
Registered User
 
Join Date: Feb 2008
Location: Huntsville, AL
Posts: 2,812
My Ride: 2000 328i
Tensor Processing Unit 2.0 was announced at Google I/O. Will offer cloud based training and inference as a service on massively powerful hardware.

Google Rattles the Tech World With a New AI Chip for All
Google brings 45 teraflops tensor flow processors to its compute cloud


Also interesting from a few days ago, watch Google Assistant beat Siri and Alexa at understanding heavy accents:
https://www.wired.com/2017/05/ai-ass...ented-english/

Spoiler alert: There's no way you can mispronounce "Benedict Cumberbatch" and have it not understand what you meant. (including Benebatch Cumberdict)

Google claims that their word error rate is now under 5%.

I don't think any of these assistants are all that valuable yet, but Google is clearly trying to get there first (and succeeding so far).
jeffro3000 is offline   Reply With Quote
Old 05-23-2017, 07:32 AM   #208
jeffro3000
Registered User
 
Join Date: Feb 2008
Location: Huntsville, AL
Posts: 2,812
My Ride: 2000 328i
Google’s AlphaGo AI defeats world Go number one Ke Jie

AlphaGo at it again.
jeffro3000 is offline   Reply With Quote
Old 05-25-2017, 10:04 AM   #209
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
^ Came to post that. This analogy isn't far off
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 07-13-2017, 05:36 AM   #210
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3

__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Last edited by cowmoo32; 07-13-2017 at 05:39 AM.
cowmoo32 is offline   Reply With Quote
Old 07-28-2017, 06:20 AM   #211
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
This company is developing processors for machine learning. They have some pictures of covnets that tell you absolutely nothing about the network but they look cool
https://www.graphcore.ai/





__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 08-12-2017, 02:41 PM   #212
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Agents that imagine and plan
https://deepmind.com/blog/agents-imagine-and-plan/

Article on the research: https://www.independent.co.uk/life-s...-a7863061.html
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 08-31-2017, 06:39 AM   #213
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
A bot has learned how to create realistic product reviews.

https://www.businessinsider.com.au/r...ke-news-2017-8
Quote:
Here's one example of a synthesised review: "I love this place. I went with my brother and we had the vegetarian pasta and it was delicious. The beer was good and the service was amazing. I would definitely recommend this place to anyone looking for a great place to go for a great breakfast and a small spot with a great deal."
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 08-31-2017, 07:00 AM   #214
325seeeye
Core Fanatic
 
Join Date: Jan 2003
Location: Rocket City
Posts: 1,169
My Ride: 325Ci, e36 325is
Quote:
Originally Posted by cowmoo32 View Post
A bot has learned how to create realistic product reviews.

https://www.businessinsider.com.au/r...ke-news-2017-8
First world people. We are striving to create neural networks which can simulate human thinking. And someone wants to use it to boost their rating on urbanspoon.
325seeeye is offline   Reply With Quote
Old 10-27-2017, 11:36 AM   #215
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
CAPTCHAs have been solved using a new type of recurrent network. Impressive on its own, but the takeaway is its ability to generalize based on a small training set. It doesn't need millions of hand-picked training examples to learn, only a relatively small set. It can then use that "knowledge" to solve CAPTCHAs.

http://science.sciencemag.org/conten...cience.aag2612
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Last edited by cowmoo32; 10-27-2017 at 11:36 AM.
cowmoo32 is offline   Reply With Quote
Old 02-26-2018, 06:54 PM   #216
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
They could have given it less-weird voice, but wow
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Old 02-27-2018, 08:51 AM   #217
tks
Registered User
 
Join Date: May 2010
Location: Online
Posts: 1,335
My Ride: slower than yours
damn its like that resident evil 1 underground little girl that took control of the entire facility
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
tks is offline   Reply With Quote
Old 02-27-2018, 09:29 AM   #218
SamDoe1
Registered User
 
Join Date: Oct 2010
Location: Minnesnowta
Posts: 3,677
My Ride: Economy Hatchback
We talked before about how AI needs to start showing rationale for the decisions it makes...I read an article last night that I can't for the life of me find now that does exactly that. Walks a user through the decision tree of why it did what it did.
SamDoe1 is offline   Reply With Quote
Old 02-27-2018, 11:39 AM   #219
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Quote:
Originally Posted by SamDoe1 View Post
We talked before about how AI needs to start showing rationale for the decisions it makes...I read an article last night that I can't for the life of me find now that does exactly that. Walks a user through the decision tree of why it did what it did.
Decision trees have been around for a looong time but they're good in certain applications and require like zero computational power. The final decision can easily be traced back up the tree. When we talked about how a network arrived at a solution, I was referring to an artificial neural network or a convolutional network, both of which have black magic, aka crazy linear algebra and vector calculus, going on behind the scenes and might be adjusting on the order of 100M different weights - a 'small' covnet I've been working with has at least 110M weights, assuming I didn't fvck up the math to calculate that.

How these networks decide to classify an image is what's 'mysterious' about them, for lack of a better term. You can see what I mean here. Some neurons seem to be picking up on faces, while others are looking at wrinkles in his clothes. As to why it decided those traits were important, we don't know. That's further evidenced by looking at an image that maximizes neuron activations (the trippy looking pictures with abstract subjects). You would think a picture of a dog maximizes the neurons responsible for classifying a dog, but instead you get some nightmarish dog-looking thing. There are no analytical solutions for the best network for a given problem, because we don't know what the best solution looks like or how to even get there. The whole field is still people brute forcing parameters and seeing what works. Obviously, the experts out there have an intuition for what will work for a given problem, but we're not at a point where we can say "We need a machine to classify brands of cell phones, here's the optimal architecture and here's a list of the optimal hyperparameters"

__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Last edited by cowmoo32; 02-27-2018 at 11:44 AM.
cowmoo32 is offline   Reply With Quote
Old 03-28-2018, 11:29 AM   #220
cowmoo32
.--. . -. .. ...
 
cowmoo32's Avatar
 
Join Date: Jul 2003
Location: HSV
Posts: 5,545
My Ride: Yukon, E92 M3
Well, speech synthesis is at a near human level now
https://research.googleblog.com/2018...esis-with.html
__________________

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
cowmoo32 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is On
Censor is ON



All times are GMT -5. The time now is 03:15 AM.


Powered by vBulletin® Version 3.8.8
Copyright ©2000 - 2020, vBulletin Solutions, Inc.
vBulletin Security provided by vBSecurity v2.2.2 (Pro) - vBulletin Mods & Addons Copyright © 2020 DragonByte Technologies Ltd.
(c) 1999 - VerticalScope Inc. All rights reserved.